Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# Re: st: Test for stationarity/non-stationarity and optimal lag structure

 From Muhammad Anees To statalist@hsphsun2.harvard.edu Subject Re: st: Test for stationarity/non-stationarity and optimal lag structure Date Thu, 19 Apr 2012 22:11:54 +0500

```You can test using any of -dfuller-, -kpss-, -dfgls-. Lag order
selection could be implement in a trial and error mode in -dfuller and
once confirmed you can use other tests with the same lag length.

When you run -dfgls variable- or -dfgls variable, maxlag(number)- you
can see that it will regress automatically. More details on unit root
tests can be read from any introductory level book on time series
analysis.

Hope this helps
Anees

On Thu, Apr 19, 2012 at 10:03 PM, Prakash Singh <prakashbhu@gmail.com> wrote:
> Dear Johns
> If you carefully look at the econometric side of test you will find
> that the variable for which you are testing dfgls is the dependent
> variable some how and then lag order is basically the lags of in
> predictor side of that equation basically taking care of
> autocorrelation.
> if the results tells you that the series is non stationary then it
> means you have to test first differenced series for stationarity.
>
>
> Prakash
>
> On Thu, Apr 19, 2012 at 2:16 AM, Johns Waldow <JohnsW@gmx.de> wrote:
>> Stata/MP 11.1 Windows
>>
>> Hello everybody,
>>
>>
>> I'm a Stata beginner. I have an time series regression with multiple
>> predictors (8, including lags of the dependent variable). I need to test for
>> stationarity and the optimal lag structure. Naturally I start with testing
>> for stationarity. If I type -- dfgls inf - I receive optimal lag orders
>> according to Ng-Perron, SC and MAIC with regards to inf and the DF-GLS Test
>> Statistic. I'm wondering how Stata can give me the optimal lag order if
>> Stata doesn't know the dependent variable.
>>
>> 1. Shouldn't I tell Stata what my dependent variable is so that I can get
>> the optimal lag order with regards to my dependent variable using - dfgls
>> [varname] --?
>> 2. When I use -- dfgls inf - Stata gives me lag orders which are not even
>> stationary. Does that mean I can use the suggested lag order but I have to
>> detrend/differentiate the series once more? If so, than again how can Stata
>> give me the lag order without knowing about my dependent variable?
>>
>> Thank you so much!
>>
>> Best regards
>>
>> Johns Waldow
>>
>> *
>> *   For searches and help try:
>> *   http://www.stata.com/help.cgi?search
>> *   http://www.stata.com/support/statalist/faq
>> *   http://www.ats.ucla.edu/stat/stata/
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/

--

Best
---------------------------
Assistant Professor/Programme Coordinator
COMSATS Institute of Information Technology
Attock 43600, Pakistan
http://www.aneconomist.com

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```